Renmin Zhujiang (Jan 2020)
Study on the Multi-parameter Inversion of Reservoir Water Quality Based on GF-1 WFV Image and Neural Network Model
Abstract
This paper establishes a multi-parameter quantitative inversion model of water qualityin multispectral remote sensing images by domestic GF-1 WFV image and neural network model toachieve high-efficiency, large-scale, continuous-space and multi-parameter change monitoring ofreservoir water quality, explores the application feasibility of domestic satellite image inremotesensing inversion of water quality, and provide technical support for lake chief system and lakeeutrophication assessment. Taking a medium-sized reservoir in Foshan City, Guangdong Province asan example, based on the GF-1 WFV image, a quantitative inversion model between 5 water qualityparameters of Chl-a, SD, TP, TN and CODMn) and image data of Dongfeng Reservoiris established withneural network models, the determination coefficient (R2) between the predicted and measuredvalues of the 5 water quality parameters all reached above 0.8, with the average relative errorsof less than 40%. The results confirm the feasibility of domestic satellite image for remotesensing inversion of water quality,which can provide a reference for the water quality andeutrophication monitoring of the reservoir.
